What is Opus 4.7?
Claude Opus 4.7 is an AI model by Anthropic. In May 2026, Nathaniel Whittemore reported Opus 4.7 scored 69.3% on Terminal Bench 2.0 and 80.5% on SweBench multilingual, comparable to GPT-5.5’s 77.8%. Whittemore also noted a competing model performed comparably to Opus 4.7 and GPT-5.5 across settings, but at 10 to 60x lower cost. Greg Isenberg said GPT-5.5 is more efficient with tool calls than Opus 4.7, which consumes more tokens for agent tasks. In August 2026, Whittemore said Opus 4.7 shipped with a new tokenizer that, per Anthropic’s documentation, produces roughly 30% more tokens for the same text; independent analyses of over a million requests found native token counts grew by 32% to 45%.
In the discourse
Attributed discussion of Opus 4.7.
When Anthropic shipped Opus 4.7 with a new tokenizer, independent analyses of over 1 million requests found token counts rose 32 to 45 percent, with real-world bills growing 12 to 27 percent (partially offset by caching), while the per-token price sheet stayed unchanged.
“The model was shipped with a new tokenizer that produced by Anthropic's own documentation, they didn't hide it, roughly 30% more tokens for the same text. So, there were quite a few independent analysis of over a million requests that found native tokens. They grew and the count grew by about 32 all the way to 45% and the real world bills grew by 12 to 27% because some of the difference was absorbed by caching.”Nathaniel Whittemore · 4 Aug 2026
Cursor's Composer 2.5 scores 69.3% on Terminal Bench 2.0 and is comparable to Opus 4.7 (80.5%) and GPT-5.5 (77.8%) on SweBench multilingual.
“It scored 69.3% on Terminal Bench 2.0, which is just behind Opus 4.7 On SweBench multilingual, it scored comparable to both Opus 4.7 at 80.5% and GPT-5.5 at 77.8%.”Nathaniel Whittemore · 20 May 2026
Cursor Composer 2.5 matched or approached Opus 4.7 and GPT 5.5 performance at 10 to 60x lower cost, making it a significant efficiency play for enterprise coding workloads.
“It performed at a very comparable level to Opus 4.7 and GPT 5.5, a little ahead of them on their medium settings and a little below them on extra high and max settings, but it did all that at 10 to 60x lower cost.”Nathaniel Whittemore · 24 May 2026
Cursor Composer 2.5 achieved comparable performance to Opus 4.7 and GPT 5.5 at 10 to 60x lower cost.
“It performed at a very comparable level to Opus 4.7 and GPT 5.5, a little ahead of them on their medium settings and a little below them on extra high and max settings, but it did all that at 10 to 60x lower cost.”Nathaniel Whittemore · 24 May 2026
GPT-5.5 is flagged as the current best model for agent tool calls, cited as significantly more token-efficient than Anthropic Opus 4.7 in agentic workflows.
“Today, by far the best model to use for something like a Hermes agent or an OpenClaw is GPT 5.5. it's so efficient with the tool calls. It doesn't eat through tokens like Opus 4.7 from Enthropic does.”Greg Isenberg · 12 May 2026